LLMStinger: Jailbreaking LLMs using RL fine-tuned LLMs

We introduce LLMStinger, a novel approach that leverages Large Language Models (LLMs) to automatically generate adversarial suffixes for jailbreak attacks. Unlike traditional methods, which require complex prompt engineering or white-box access, LLMStinger uses a reinforcement learning (RL) loop to fine-tune an attacker LLM, generating new suffixes based on existing attacks for harmful questions from the HarmBench benchmark. Our method significantly outperforms existing red-teaming approaches (we compared against 15 of the latest methods), achieving a +57.2% improvement in Attack Success Rate (ASR) on LLaMA2-7B-chat and a +50.3% ASR increase on Claude 2, both models known for their extensive safety measures. Additionally, we achieved a 94.97% ASR on GPT-3.5 and 99.4% on Gemma-2B-it, demonstrating the robustness and adaptability of LLMStinger across open and closed-source models.

Survey of Cultural Awareness in Language Models: Text and Beyond

Large-scale deployment of large language models (LLMs) in various applications, such as chatbots and virtual assistants, requires LLMs to be culturally sensitive to the user to ensure inclusivity. Culture has been widely studied in psychology and anthropology, and there has been a recent surge in research on making LLMs more culturally inclusive in LLMs that goes beyond multilinguality and builds on findings from psychology and anthropology. In this paper, we survey efforts towards incorporating cultural awareness into text-based and multimodal LLMs. We start by defining cultural awareness in LLMs, taking the definitions of culture from anthropology and psychology as a point of departure. We then examine methodologies adopted for creating cross-cultural datasets, strategies for cultural inclusion in downstream tasks, and methodologies that have been used for benchmarking cultural awareness in LLMs. Further, we discuss the ethical implications of cultural alignment, the role of Human-Computer Interaction in driving cultural inclusion in LLMs, and the role of cultural alignment in driving social science research. We finally provide pointers to future research based on our findings about gaps in the literature.

Revealing Fine-Grained Values and Opinions in Large Language Models

Uncovering latent values and opinions in large language models (LLMs) can help identify biases and mitigate potential harm. Recently, this has been approached by presenting LLMs with survey questions and quantifying their stances towards morally and politically charged statements. However, the stances generated by LLMs can vary greatly depending on how they are prompted, and there are many ways to argue for or against a given position. In this work, we propose to address this by analysing a large and robust dataset of 156k LLM responses to the 62 propositions of the Political Compass Test (PCT) generated by 6 LLMs using 420 prompt variations. We perform coarse-grained analysis of their generated stances and fine-grained analysis of the plain text justifications for those stances. For fine-grained analysis, we propose to identify tropes in the responses: semantically similar phrases that are recurrent and consistent across different prompts, revealing patterns in the text that a given LLM is prone to produce. We find that demographic features added to prompts significantly affect outcomes on the PCT, reflecting bias, as well as disparities between the results of tests when eliciting closed-form vs. open domain responses. Additionally, patterns in the plain text rationales via tropes show that similar justifications are repeatedly generated across models and prompts even with disparate stances.

RLSF: Reinforcement Learning via Symbolic Feedback

In recent years, large language models (LLMs) have had a dramatic impact on various sub-fields of AI, most notably on natural language understanding tasks. However, there is widespread agreement that the logical reasoning capabilities of contemporary LLMs are, at best, fragmentary (i.e., may work well on some problem instances but fail dramatically on others). While traditional LLM fine-tuning approaches (e.g., those that use human feedback) do address this problem to some degree, they suffer from many issues, including unsound black-box reward models, difficulties in collecting preference data, and sparse scalar reward values. To address these challenges, we propose a new training/fine-tuning paradigm we refer to as Reinforcement Learning via Symbolic Feedback (RLSF), which is aimed at enhancing the reasoning capabilities of LLMs. In the RLSF setting, the LLM that is being trained/fine-tuned is considered as the RL agent, while the environment is allowed access to reasoning or domain knowledge tools (e.g., solvers, algebra systems). Crucially, in RLSF, these reasoning tools can provide feedback to the LLMs via poly-sized certificates (e.g., proofs), that characterize errors in the LLM-generated object with respect to some correctness specification. The ability of RLSF-based training/fine-tuning to leverage certificate-generating symbolic tools enables sound fine-grained (token-level) reward signals to LLMs, and thus addresses the limitations of traditional reward models mentioned above. Via extensive evaluations, we show that our RLSF-based fine-tuning of LLMs outperforms traditional approaches on two different applications, namely, program synthesis from natural language pseudo-code to programming language (C++) and solving the Game of 24.

Overview of the 2023 ICON Shared Task on Gendered Abuse Detection in Indic Languages

This paper reports the findings of the ICON 2023 on Gendered Abuse Detection in Indic Languages. The shared task deals with the detection of gendered abuse in online text. The shared task was conducted as a part of ICON 2023, based on a novel dataset in Hindi, Tamil and the Indian dialect of English. The participants were given three subtasks with the train dataset consisting of approximately 6500 posts sourced from Twitter. For the test set, approximately 1200 posts were provided. The shared task received a total of 9 registrations. The best F-1 scores are 0.616 for subtask 1, 0.572 for subtask 2 and, 0.616 and 0.582 for subtask 3. The paper contains examples of hateful content owing to its topic.

Factcheck-GPT: End-to-End Fine-Grained Document-Level Fact-Checking and Correction of LLM Output

The increased use of large language models (LLMs) across a variety of real-world applications calls for mechanisms to verify the factual accuracy of their outputs. In this work, we present a holistic end-to-end solution for annotating the factuality of LLM-generated responses, which encompasses a multi-stage annotation scheme designed to yield detailed labels concerning the verifiability and factual inconsistencies found in LLM outputs. We design and build an annotation tool to speed up the labelling procedure and ease the workload of raters. It allows flexible incorporation of automatic results in any stage, e.g. automatically-retrieved evidence. We further construct an open-domain document-level factuality benchmark in three-level granularity: claim, sentence and document. Preliminary experiments show that FacTool, FactScore and Perplexity.ai are struggling to identify false claims with the best F1=0.53. Annotation tool, benchmark and code are available at https://github.com/yuxiaw/Factcheck-GPT.

The Uli Dataset: An Exercise in Experience Led Annotation of oGBV

Online gender based violence has grown concomitantly with adoption of the internet and social media. Its effects are worse in the Global majority where many users use social media in languages other than English. The scale and volume of conversations on the internet has necessitated the need for automated detection of hate speech, and more specifically gendered abuse. There is, however, a lack of language specific and contextual data to build such automated tools. In this paper we present a dataset on gendered abuse in three languages- Hindi, Tamil and Indian English. The dataset comprises of tweets annotated along three questions pertaining to the experience of gender abuse, by experts who identify as women or a member of the LGBTQIA community in South Asia. Through this dataset we demonstrate a participatory approach to creating datasets that drive AI systems.

Why Should This Article Be Deleted? Transparent Stance Detection in Multilingual Wikipedia Editor Discussions

The moderation of content on online platforms is usually non-transparent. On Wikipedia, however, this discussion is carried out publicly and the editors are encouraged to use the content moderation policies as explanations for making moderation decisions. Currently, only a few comments explicitly mention those policies -- 20% of the English ones, but as few as 2% of the German and Turkish comments. To aid in this process of understanding how content is moderated, we construct a novel multilingual dataset of Wikipedia editor discussions along with their reasoning in three languages. The dataset contains the stances of the editors (keep, delete, merge, comment), along with the stated reason, and a content moderation policy, for each edit decision. We demonstrate that stance and corresponding reason (policy) can be predicted jointly with a high degree of accuracy, adding transparency to the decision-making process. We release both our joint prediction models and the multilingual content moderation dataset for further research on automated transparent content moderation.

Topic-Guided Sampling For Data-Efficient Multi-Domain Stance Detection

Stance Detection is concerned with identifying the attitudes expressed by an author towards a target of interest. This task spans a variety of domains ranging from social media opinion identification to detecting the stance for a legal claim. However, the framing of the task varies within these domains, in terms of the data collection protocol, the label dictionary and the number of available annotations. Furthermore, these stance annotations are significantly imbalanced on a per-topic and inter-topic basis. These make multi-domain stance detection a challenging task, requiring standardization and domain adaptation. To overcome this challenge, we propose $\textbf{T}$opic $\textbf{E}$fficient $\textbf{St}$anc$\textbf{E}$ $\textbf{D}$etection (TESTED), consisting of a topic-guided diversity sampling technique and a contrastive objective that is used for fine-tuning a stance classifier. We evaluate the method on an existing benchmark of $16$ datasets with in-domain, i.e. all topics seen and out-of-domain, i.e. unseen topics, experiments. The results show that our method outperforms the state-of-the-art with an average of $3.5$ F1 points increase in-domain, and is more generalizable with an averaged increase of $10.2$ F1 on out-of-domain evaluation while using $\leq10\%$ of the training data. We show that our sampling technique mitigates both inter- and per-topic class imbalances. Finally, our analysis demonstrates that the contrastive learning objective allows the model a more pronounced segmentation of samples with varying labels.

Thorny Roses: Investigating the Dual Use Dilemma in Natural Language Processing

Dual use, the intentional, harmful reuse of technology and scientific artefacts, is a problem yet to be well-defined within the context of Natural Language Processing (NLP). However, as NLP technologies continue to advance and become increasingly widespread in society, their inner workings have become increasingly opaque. Therefore, understanding dual use concerns and potential ways of limiting them is critical to minimising the potential harms of research and development. In this paper, we conduct a survey of NLP researchers and practitioners to understand the depth and their perspective of the problem as well as to assess existing available support. Based on the results of our survey, we offer a definition of dual use that is tailored to the needs of the NLP community. The survey revealed that a majority of researchers are concerned about the potential dual use of their research but only take limited action toward it. In light of the survey results, we discuss the current state and potential means for mitigating dual use in NLP and propose a checklist that can be integrated into existing conference ethics-frameworks, e.g., the ACL ethics checklist.